Benchmark explainer
What is HLE?
Frontier expert knowledge/reasoning across 100+ subjects
Humanity's Last Exam — Frontier expert knowledge/reasoning across 100+ subjects; scored as Accuracy + calibration/confidence-error.
How it's scored
- Metric
- Accuracy + calibration/confidence-error
- Score ceiling
- 100
- Construction
- ~2500 Qs from ~1000 expert contributors, multi-stage review; ~10-14% multimodal
- Human baseline
- No aggregate (at/beyond individual expert frontier)
How trustworthy
Discrimination
100/100Does it still separate models?
Saturation headroom
98.7/100How far from ceiling / clustered at the top?
Contamination resistance
14.3/100Public vs held-out; training-leak risk.
Harness comparability
100/100Is it apples-to-apples, or mixed / vendor-optimized?
Freshness
85.3/100How old is the benchmark?
Contamination history: Public + private holdout to detect gaming; public leakage expected
Limitations: Short-answer/MC lucky hits; obscure recall; public portion contaminates; not agentic
Who leads HLE
| Model | Score | Evidence |
|---|---|---|
| GPT-6 Astra | 52.52 | unverified |
| Claude Fable 5.1 | 43.8 | unverified |
| Gemini 3.1 Pro | 43.74 | unverified |
| Claude Sonnet 5.5 | 43.4 | unverified |
| Gemini 3.8 Flash | 41.72 | unverified |
Compare the top HLE scorers
Frequently asked questions
HLE (Humanity's Last Exam): Humanity's Last Exam — Frontier expert knowledge/reasoning across 100+ subjects; scored as Accuracy + calibration/confidence-error.
GPT-6 Astra leads HLE at 52.52. The full leaderboard above lists every recorded measurement, not just the headline number.
42 models have recorded scores on HLE, spanning a score spread of 52.52.
tensor.news grades HLE A for integrity (score 85/100), ranking #34 of 65 benchmarks we assess across discrimination, saturation, contamination resistance, harness comparability, and freshness.
No — HLE still has headroom and continues to discriminate between models rather than bunching them at the ceiling.
Its test set is public, so contamination risk is on the table: a high score may partly reflect training-data overlap rather than capability.
Scores are reported under a consistent harness, so comparisons on HLE are reasonably apples-to-apples.
Every HLE measurement is source-backed and tagged reproduced, self-reported, or unverified; the underlying record cites agi.safe.ai.